A hoisting risk early warning method and system for hoisting a drone

Abstract

The present application relates to the technical field of safety early warning, and more particularly to a hoisting risk early warning method and system for hoisting unmanned aerial vehicles, comprising analyzing the real-time stress state and stability of hoisted articles and evaluating the risk level of hoisted articles; dynamically calculating the weight coefficients of personnel risk, equipment risk and hoisted article risk; performing weighted fusion based on the determined dynamic weight coefficients to generate a comprehensive risk index; the present application introduces a dynamic weight calculation mechanism based on multi-dimensional real-time cost factors, quantitatively fuses safety cost, article value cost, equipment wear cost, energy consumption cost and time cost, and adjusts the weight proportion of different risk categories in real time accordingly; the scheme enables the risk perception center of gravity of the entire early warning system to intelligently migrate with the work scene, and is more conservative in harsh environments to ensure safety, and pays more attention to the integrity when hoisting valuable articles, thereby realizing intelligent self-adaptation and flexible decision-making of the risk evaluation strategy.

Description

Technical Field

[0001] This invention relates to the field of safety early warning technology, and in particular to a method and system for early warning of hoisting risks for hoisting drones. Background Technology

[0002] During drone hoisting, the stability of the hoisted items is directly related to operational safety. If risks such as uneven cable tension, instability of the hoisted items, or collisions occur, it may not only cause damage to the hoisted items and malfunction of the drone equipment, but also threaten the safety of surrounding personnel. Therefore, building a precise and efficient hoisting risk early warning system has become one of the core requirements for the development of drone hoisting technology.

[0003] Current risk warning technologies in the field of drone hoisting still have many shortcomings and are insufficient to meet the safety requirements of complex operation scenarios. Existing technologies mostly use fixed weights in risk fusion assessment, which cannot be dynamically adjusted according to the operating environment, equipment status, and task requirements, resulting in a low degree of matching between the assessment results and the actual scenario. At the same time, most existing technologies can only achieve real-time risk warnings, lacking the ability to trace the causes of risks and predict potential future risks. Operators can only passively deal with risks that have already occurred, making it difficult to form an effective risk prevention and control loop.

[0004] To address the shortcomings of the existing technologies, this invention proposes a method and system for early warning of hoisting risks for drones. It constructs a dynamic weighting mechanism driven by multiple cost factors to improve the accuracy of risk assessment, introduces causal discovery algorithms and dynamic Bayesian networks to achieve risk root cause tracing and future prediction, and finally outputs targeted control and adjustment suggestions through a hierarchical early warning mechanism that combines real-time risk index with future prediction results, forming a complete risk prevention and control closed loop, effectively improving the safety and reliability of drone hoisting operations. Summary of the Invention

[0005] To overcome the problems mentioned in the background art, the present invention proposes a method and system for early warning of hoisting risks for hoisting drones.

[0006] The technical solution of this invention is: a method for early warning of lifting risks for hoisting drones, comprising the following steps:

[0007] S1: Real-time monitoring of the force data of multiple sets of hoisting cables and the attitude data of the hoisted items in the hoisting system;

[0008] S2: Based on the acquired force and attitude data, analyze the real-time force state and stability of the hoisted item, and assess the risk level of the hoisted item;

[0009] S3: Based on the real-time operating environment and task parameters, dynamically calculate the weighting coefficients of personnel risk, equipment risk and hoisting item risk. The calculation basis of the weighting coefficients includes safety cost, hoisting item cost, equipment wear and tear cost, energy cost and time cost.

[0010] S4: Combining the risk assessment results of the hoisted items, and comprehensively assessing personnel and equipment risks, a comprehensive risk index is generated by weighting and integrating the results based on the determined dynamic weight coefficients.

[0011] S5: Based on historical monitoring data, control command sequences, and risk event records, a risk causal graph is constructed using a causal discovery algorithm to infer the root cause of the current risk status;

[0012] S6: Using the obtained risk cause-effect diagram, extrapolate the preset control program and real-time input control commands to predict potential risks in the future time period;

[0013] S7: Based on the comprehensive risk index and future risk prediction results, trigger the corresponding level of risk warning signal, and output control adjustment suggestions according to the warning level.

[0014] Preferably, real-time monitoring of the stress data of multiple sets of lifting cables and the attitude data of the lifted items in the lifting system specifically includes:

[0015] S11: Force sensors are installed on each hoisting cable to synchronously collect tension data of each cable at a sampling frequency of 100Hz.

[0016] S12: Install an inertial measurement unit on the hoisted item to collect the acceleration, angular velocity and attitude angle data of the hoisted item in real time;

[0017] S13: By installing a ranging module on the drone, the relative position and distance between the hoisted object and the drone body are monitored.

[0018] As a preferred option, when assessing the risk level of hoisted items, the following should be included:

[0019] S21: Based on the real-time force data of each hoisting cable, calculate the resultant force, resultant moment and force imbalance parameters of the hoisted item;

[0020] S22: Based on the acquired attitude data of the hoisted item, analyze the vibration characteristics, attitude stability and torsional angle parameters of the hoisted item;

[0021] S23: Couple the stress imbalance parameter with the attitude stability parameter to identify the instability mode of the hoisted item;

[0022] S24: Based on the instability mode, a risk quantification model is used to calculate the risk factors of the hoisted items;

[0023] S25: Based on the material properties, structural characteristics, and current environmental conditions of the hoisted items, dynamically adjust the risk level threshold range, and determine the risk factors of the hoisted items according to the risk level threshold range to determine the specific risk level.

[0024] Preferably, when coupling the stress imbalance parameter with the attitude stability parameter to identify the instability mode of the hoisted object, the specific steps include:

[0025] S231: Combine the force imbalance parameter, the reciprocal of the attitude stability parameter, the critical vibration energy ratio parameter, and the torsion angle parameter to construct a coupled analysis feature vector;

[0026] S232: The matching degree of the coupling analysis feature vector is calculated with a set of preset instability mode feature vectors respectively. The matching degree calculation represents the similarity between the current state and each preset instability mode. The preset instability modes include translational instability mode dominated by force imbalance, rotational instability mode dominated by moment imbalance, structural vibration instability mode caused by resonance, and torsional instability mode caused by torsional accumulation.

[0027] S233: Based on the calculated matching degree, determine the instability mode with the highest matching degree to the current state and identify it as the current main instability mode.

[0028] As a preferred method, when dynamically calculating the weighting coefficients for personnel risk, equipment risk, and hoisted item risk based on the real-time operating environment and task parameters, the specific calculations include:

[0029] S31: Real-time acquisition of environmental parameters for the current operation, task attribute parameters for the hoisting task, status parameters of the UAV equipment, and task constraint parameters from external systems;

[0030] S32: Calculate the safety cost factor, hoisting cost factor, equipment loss cost factor, energy consumption cost factor, and time cost factor based on various parameters.

[0031] S33: Convert the safety cost factor, hoisting item cost factor, equipment wear and tear cost factor, energy consumption cost factor, and time cost factor into the initial values ​​of personnel risk weight coefficient, equipment risk weight coefficient, and hoisting item risk weight coefficient through a predetermined mapping rule;

[0032] S34: Normalize the initial values ​​of personnel risk weight coefficient, equipment risk weight coefficient, and hoisted item risk weight coefficient, and output them as the final dynamic weight coefficients.

[0033] As a preferred option, the predetermined mapping rule is:

[0034] The safety cost factor is mainly used to determine the initial value of the personnel risk weight coefficient, and the two are positively correlated. The hoisting item cost factor is mainly used to determine the initial value of the hoisting item risk weight coefficient, and the two are positively correlated. The equipment loss cost factor and the energy consumption cost factor are used together to determine the initial value of the equipment risk weight coefficient, and the initial value of the equipment risk weight coefficient is the result of the weighted sum of the two cost factors. The time cost factor is used to globally adjust the initial values ​​of the personnel risk weight coefficient, the equipment risk weight coefficient, and the hoisting item risk weight coefficient. When the time cost factor increases, the values ​​of all risk weight coefficients are reduced simultaneously while keeping the weight ratio of the three factors basically unchanged.

[0035] As a preferred approach, when constructing a risk causal graph using a causal discovery algorithm based on historical monitoring data, control command sequences, and risk event records to infer the root cause of the current risk state, the specific methods include:

[0036] S51: Collect multi-source time-series data within a historical time window, including sensor data sequences, control command sequences issued by the UAV flight control system, and recorded risk event label sequences;

[0037] S52: Based on multi-source time series data, a causal discovery algorithm is used to analyze and determine the causal relationships between various data variables;

[0038] S53: Based on the established causal relationships, construct a directed graph model as a risk causal graph, where the nodes of the graph represent the variables of the multi-source time series data, and the directed edges between the nodes represent the direction of the causal relationship between the variables.

[0039] S54: Based on the constructed risk causal graph, combined with the current real-time monitoring data and status, infer the root cause node and key causal path that caused the current risk status.

[0040] As a preferred approach, when inferring the root cause nodes and key causal paths leading to the current risk state based on the constructed risk cause-effect graph and combined with current real-time monitoring data and status, the specific steps include:

[0041] S541: Evidence that maps current real-time monitoring data and system status to corresponding observed variable nodes and control variable nodes in a risk causal graph;

[0042] S542: In the risk causal graph, starting from the triggered risk variable node, trace back along the directed edge to identify all possible predecessor nodes as candidate cause nodes.

[0043] S543: Based on the causal strength and the degree to which real-time data deviates from the normal threshold, calculate the contribution of each candidate cause node to the current risk state;

[0044] S544: Based on the contribution ranking, the top-ranked candidate cause nodes and the paths connecting them to risk nodes are presumed to be the root cause and key causal path of the current risk.

[0045] Preferably, when using the obtained risk cause-effect diagram to extrapolate the preset control program and real-time input control commands to predict potential risks in the future time period, the specific steps include:

[0046] S61: Analyze the control programs that are currently to be executed and those that will be executed in the future, and capture the control instructions that are input in real time, and convert them into the input sequence of the corresponding control variable node in the risk cause-effect graph;

[0047] S62: Based on the risk cause-effect diagram and combined with the current system state, extrapolate the impact of the control input sequence in the future time period along the direction of the causal relationship in the cause-effect diagram;

[0048] S63: Based on the simulation results, predict the probability of a specific risk event occurring in the future time period and output the risk prediction results.

[0049] A hoisting risk early warning system for hoisting drones includes:

[0050] The data acquisition module is used to monitor the stress data of multiple sets of hoisting cables and the attitude data of the hoisted items in the hoisting system in real time.

[0051] The risk analysis module for hoisted items is used to analyze the real-time stress state and stability of hoisted items based on the acquired stress and attitude data, and to assess the risk level of the hoisted items.

[0052] The dynamic weight calculation module is used to dynamically calculate the weight coefficients of personnel risk, equipment risk, and hoisted item risk based on the real-time operating environment and task parameters.

[0053] The comprehensive risk assessment module is used to integrate independently assessed personnel and equipment risks. It performs weighted fusion based on the dynamic weight coefficients determined by the dynamic weight calculation module to generate a comprehensive risk index.

[0054] The causal analysis and root cause inference module is used to construct a risk causal graph based on historical monitoring data, control command sequences and risk event records, and infer the root cause of the current risk status.

[0055] The risk prediction module is used to use a risk cause-effect diagram to extrapolate from preset control programs or real-time input control commands and predict potential risks in the future time period.

[0056] The early warning and response module is used to trigger risk warning signals of corresponding levels based on the comprehensive risk index and future risk prediction results, and output control adjustment suggestions according to the warning level.

[0057] The beneficial effects of this invention are:

[0058] 1. Compared to existing technologies that typically assign fixed weights to personnel, equipment, and hoisted items during comprehensive risk assessment, resulting in a rigid and inflexible approach that cannot adapt to the dynamic priorities of safety, efficiency, and economic costs in hoisting tasks, this invention introduces a dynamic weighting mechanism based on multi-dimensional real-time cost factors. This mechanism quantifies and integrates safety costs, item value costs, equipment wear and tear costs, energy costs, and time costs, and adjusts the weighting of different risk categories in real time accordingly. This solution enables the risk perception focus of the entire early warning system to intelligently shift with the operational scenario, becoming more conservative in harsh environments to ensure safety, and focusing more on integrity when hoisting valuable items, thus achieving intelligent self-adaptation and flexible decision-making in risk assessment strategies.

[0059] 2. Compared to existing technologies that primarily rely on monitoring a single physical quantity, such as focusing only on overweight or tilt angle for threshold alarms, these solutions often only capture the surface phenomena of risks without deeply analyzing the fundamental mechanisms of instability, resulting in delayed and weak early warnings. This invention creatively constructs a high-dimensional data network that synchronously senses force, attitude, and position. By coupling and analyzing these multi-source data, it achieves accurate identification of various specific instability modes, such as translational instability, rotational instability, resonance instability, and torsional instability. This solution enables the system to see through the mechanical essence behind complex dynamics, upgrading from simple state alarms to precise mechanism diagnosis, providing a solid scientific basis for subsequent targeted interventions, and greatly improving the depth and accuracy of risk identification.

[0060] 3. Compared to existing technologies, which are mostly limited to real-time alarms and simple post-event attribution of risks, lacking in-depth tracing of risk causes and the ability to predict future risks, this invention innovatively applies a causal discovery algorithm to automatically construct a risk causal graph connecting sensor data, control commands, and risk events from historical data. This solution can not only quickly trace and lock the root cause nodes and risk-causing paths when risks occur, but also use this graph to deduce future control commands to be executed and predict potential risks in advance. This achieves a fundamental shift from passive alarm to proactive prediction, providing operators with a valuable decision-making buffer period, thereby preventing problems before they occur.

[0061] 4. Compared to existing technologies that generally use static, one-size-fits-all risk level determination thresholds, these methods cannot fully consider the actual impact of the characteristics of different hoisted items and changing environmental conditions on risk tolerance, resulting in poor adaptability. This invention proposes a context-aware dynamic threshold adjustment method. Based on multiple specific factors such as the material properties of the hoisted item (brittle or tough), the wind speed of the current working environment, and the physical dimensions of the hoisted item, the method dynamically calculates and adjusts the risk level determination threshold. This approach enables the risk warning system to possess context-aware judgment capabilities similar to those of human experts. For high-risk combinations, such as brittle materials in windy environments, it automatically applies stricter standards, significantly improving the rationality and reliability of the system's risk assessment under different complex working conditions. Attached Figure Description

[0062] Figure 1 The diagram shown is a flowchart illustrating the lifting risk warning method for lifting drones according to the present invention.

[0063] Figure 2 The diagram shown is a structural schematic of the hoisting risk warning system for hoisting drones according to the present invention. Detailed Implementation

[0064] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0065] Please see Figure 1 The present invention provides an embodiment: a method for early warning of lifting risks for hoisting drones, comprising the following steps:

[0066] Step 1: Real-time monitoring of the stress data of multiple sets of lifting cables and the attitude data of the lifted items in the lifting system, specifically including:

[0067] S11: Force sensors are installed on each hoisting cable to synchronously collect tension data of each cable at a sampling frequency of 100Hz.

[0068] S12: Install an inertial measurement unit on the hoisted item to collect the acceleration, angular velocity and attitude angle data of the hoisted item in real time;

[0069] S13: By installing a ranging module on the drone, the relative position and distance between the hoisted object and the drone body are monitored.

[0070] In this embodiment, the present invention constructs a multi-dimensional real-time data acquisition network in the hoisting system. Force sensors are deployed on each hoisting cable, and tension data of each cable are synchronously acquired at a sampling frequency of not less than 100Hz to achieve real-time perception of load distribution. Inertial measurement units are installed on the hoisted object to acquire its acceleration, angular velocity, and attitude angle data in real time, accurately capturing its motion state and dynamic behavior. Simultaneously, the relative position and distance changes between the hoisted object and the UAV are continuously monitored through a ranging module mounted on the UAV. The beneficial effect of this solution is that, through the synchronous and high-frequency acquisition of force, attitude, and displacement, a high-precision, real-time three-dimensional data foundation is provided for risk analysis, enabling the system to accurately calculate the force balance, vibration spectrum, and spatial pose of the hoisted object, providing a reliable quantitative basis for subsequent accurate assessment of the force state, identification of instability modes, and prediction of potential risks.

[0071] Step Two: Based on the acquired force and attitude data, analyze the real-time force state and stability of the hoisted object, and assess the risk level of the hoisted object. This includes:

[0072] S21: Based on the real-time force data of each hoisting cable, calculate the resultant force, resultant moment, and force imbalance parameters of the hoisted item, specifically including:

[0073] S211: In a three-dimensional coordinate system, the force vector of the i-th hoisting cable is represented as... The position vector of its point of action relative to the center of mass of the hoisted object is represented as: ;

[0074] S212: Calculate the resultant force and resultant moment on the hoisted item;

[0075] S213: Calculate the stress imbalance parameter of the hoisted item. The calculation principle formula is as follows:

[0076] ;

[0077] in, This refers to the parameter representing the unevenness of force distribution on the hoisted object. The standard deviation of the force on each cable. This represents the average value of the force on each cable. This represents the maximum stress on each cable. This represents the minimum stress on each cable. This is the rated load for a single cable. The standard deviation of the angle between the direction of each cable force and the direction of the resultant force. , and The weighting factor is determined based on the type of item being lifted and the arrangement of the lifting points.

[0078] S22: Based on the acquired attitude data of the hoisted item, analyze the vibration characteristics, attitude stability, and torsional angle parameters of the hoisted item, specifically including:

[0079] S221: Perform a short-time Fourier transform on the acceleration data of the hoisted item to extract the energy proportion of the acceleration signal within the dangerous frequency range. The calculation principle formula is as follows:

[0080] ;

[0081] in, For the spectrum of the acceleration signal, The highest frequency analyzed, This represents the lower limit of the dangerous frequency range. This represents the upper limit of the dangerous frequency range;

[0082] S222: Calculate the attitude stability parameters of the hoisted item. The calculation principle formula is as follows:

[0083] ;

[0084] in, These are the attitude stability parameters for the hoisted items. Let be the angular velocity of the hoisted object about the x-axis. Let be the angular velocity of the hoisted item about the y-axis. Let be the angular velocity of the hoisted item about the z-axis. The actual Euler angle of the hoisted object about the x-axis. The actual Euler angle of the hoisted item about the y-axis. The actual Euler angle of the hoisted item about the z-axis. The desired Euler angle of the hoisted object about the x-axis. The desired Euler angle of the hoisted object about the y-axis. The desired Euler angle of the hoisted object about the y-axis;

[0085] S223: Calculate the torsion angle parameters of the hoisted item. The calculation principle formula is as follows:

[0086] ;

[0087] in, The torsion angle parameter for the hoisted item. This is the difference between the actual yaw angle and the desired yaw angle of the hoisted item. This is the dynamic torsional coefficient, which is related to the size and shape of the object being lifted. This is the derivative of the yaw angle difference with respect to time, i.e., the torsional velocity;

[0088] S23: Couple the stress imbalance parameter with the attitude stability parameter to identify the instability mode of the hoisted object, specifically including:

[0089] S231: Combine the force imbalance parameter, the reciprocal of the attitude stability parameter, the proportion of dangerous vibration energy, and the torsion angle parameter to construct a coupled analysis feature vector;

[0090] S232: The matching degree of the coupling analysis feature vector is calculated with a set of preset instability mode feature vectors respectively. The matching degree calculation represents the similarity between the current state and each preset instability mode. The preset instability modes include translational instability mode dominated by force imbalance, rotational instability mode dominated by moment imbalance, structural vibration instability mode caused by resonance, and torsional instability mode caused by torsional accumulation.

[0091] S233: Based on the calculated matching degree, determine the instability mode with the highest matching degree to the current state and identify it as the current main instability mode;

[0092] S24: Based on instability modes, a risk quantification model is used to calculate the risk factors of hoisted items, specifically including:

[0093] S241: Normalize the parameters of force imbalance, reciprocal of attitude stability, proportion of dangerous vibration energy, torsion angle, and highest matching degree.

[0094] S242: Based on the identified main instability modes, weight coefficients are assigned to each parameter. Then, based on the parameters and weight coefficients, the risk level of the hoisted item is calculated using a monotonically increasing risk quantification function. The principle formula is as follows:

[0095] ;

[0096] in, Based on the risk level of the hoisted items, It is an exponential function. This is the normalized value of the stress imbalance parameter. These are the normalized values ​​of the attitude stability parameters. This is the normalized value of the parameter representing the proportion of hazardous vibration energy. This is the normalized value of the torsion angle parameter. For the highest matching degree, , , , and The weighting coefficients assigned;

[0097] S25: Based on the material properties, structural characteristics, and current environmental conditions of the hoisted items, dynamically adjust the risk level threshold range, and determine the risk factors of the hoisted items according to the risk level threshold range to determine the specific risk level.

[0098] When assigning weight coefficients to each parameter based on the identified main instability modes, the allocation rule is as follows:

[0099] When the main instability mode identified is translational instability dominated by force imbalance, the force imbalance parameter is assigned the largest weight coefficient.

[0100] When the main instability mode identified is rotational instability dominated by torque imbalance, the weight coefficient with the largest reciprocal of the attitude stability parameter is assigned.

[0101] When the main instability mode identified is structural vibration instability caused by resonance, the parameter with the largest weighting coefficient is assigned to the critical vibration energy ratio parameter.

[0102] When the identified primary instability mode is torsional instability caused by torsional accumulation, the torsional angle parameter is assigned the largest weight coefficient.

[0103] Among them, the highest matching degree parameter is assigned a small non-zero weight coefficient in all types of instability modes.

[0104] The specific method for dynamically adjusting the risk level threshold range is as follows:

[0105] A baseline threshold range is set. Based on the material properties of the hoisted item, a material property influence factor is determined, where brittle materials correspond to a positive influence factor, and tough materials correspond to a zero influence factor. A wind speed influence factor is determined based on the ratio of the current ambient wind speed to a preset wind speed threshold. A size influence factor is determined based on the ratio of the hoisted item's size to a preset size threshold. The baseline threshold range is then calculated using the material property influence factor, wind speed influence factor, and size influence factor to obtain the adjusted threshold range. The underlying formula is as follows:

[0106] ;

[0107] ;

[0108] in, The lower limit threshold of the baseline risk level, The baseline risk level upper limit is the maximum risk level threshold preset under standard conditions. This is the adjusted lower limit threshold for the risk level. This refers to the adjusted upper limit threshold for risk levels. The brittleness coefficient of the material. The wind speed influence coefficient, This is the size influence coefficient.

[0109] In this embodiment, the present invention calculates the resultant force and torque of the cable and innovatively combines the standard deviation, range, and directional angle deviation of the force to construct a force imbalance parameter that comprehensively reflects the force distribution. Simultaneously, it performs spectral and temporal analysis on the attitude data to extract multi-dimensional features such as dangerous vibration energy, attitude stability deviation, and dynamic torsional angle. Furthermore, these feature parameters are combined into a feature vector, and by performing similarity matching with preset typical instability modes such as force imbalance, torque imbalance, resonance, and torsion, the currently dominant instability mode is accurately identified. Based on this, an adaptive weight allocation strategy based on instability modes is adopted to weight and fuse the normalized feature parameters, and the final risk level is quantified and calculated through a monotonically increasing risk function. Furthermore, the solution dynamically adjusts the risk level judgment threshold based on the material brittleness of the hoisted items, ambient wind speed, and physical dimensions, enhancing the adaptability and accuracy of the assessment. Through the fusion and coupling analysis of multi-source data, it achieves a profound understanding of the force-motion coupling instability mechanism, overcoming the limitations of single-parameter assessment. Adaptive weight allocation based on pattern recognition enables risk assessment to focus on the most significant current hazards, improving the targeting and sensitivity of early warnings. The introduction of dynamic thresholds allows the system to make flexible judgments based on the specific work objects and environmental differences, significantly improving the reliability and practicality of risk warnings under complex and variable working conditions, providing a solid decision-making basis for subsequent precise intervention.

[0110] Step 3: Based on the real-time operating environment and task parameters, dynamically calculate the weighting coefficients for personnel risk, equipment risk, and risk of the hoisted items. The calculation of these weighting coefficients is based on factors including safety cost, hoisted item cost, equipment wear and tear cost, energy cost, and time cost. Specifically:

[0111] S31: Real-time acquisition of environmental parameters for the current operation, task attribute parameters for the hoisting task, status parameters of the UAV equipment, and task constraint parameters from external systems. Specifically, the acquired parameters are:

[0112] Current environmental parameters for the operation: wind speed, wind direction, air density, visibility, and operating height; task attribute parameters for the hoisting task: type, value, physical dimensions, structural strength, vulnerability level, and hoisting point location information of the hoisted item; status parameters for the UAV equipment: cumulative operating time of the UAV and hoisting equipment, current payload, remaining battery power, health status assessment values ​​of the motor and transmission mechanism, and predicted remaining lifespan of key components; task constraint parameters for the external system: task deadline, current progress, task criticality level, and maximum allowable energy consumption.

[0113] S32: Calculate the safety cost factor, hoisting cost factor, equipment loss cost factor, energy consumption cost factor, and time cost factor based on various parameters.

[0114] S33: Convert the safety cost factor, hoisting item cost factor, equipment wear and tear cost factor, energy consumption cost factor, and time cost factor into the initial values ​​of personnel risk weight coefficient, equipment risk weight coefficient, and hoisting item risk weight coefficient through a predetermined mapping rule;

[0115] S34: Normalize the initial values ​​of personnel risk weight coefficient, equipment risk weight coefficient, and hoisted item risk weight coefficient, and output them as the final dynamic weight coefficients.

[0116] S35: When the calculated final dynamic weight coefficient changes abruptly relative to the value of the previous calculation period, and the change exceeds the preset smoothing threshold, a first-order low-pass filter is applied to the weight coefficient of this calculation to make the weight coefficient smoothly transition to the new value, so as to avoid drastic fluctuations in risk warning decisions due to instantaneous parameter fluctuations.

[0117] The calculation methods for each cost factor are as follows:

[0118] The safety cost factor is calculated based on environmental parameters such as wind speed, operating height, and visibility, as well as task attribute parameters such as the physical size and vulnerability level of the hoisted items. The higher the wind speed, the higher the height, the lower the visibility, the larger the item size, and the higher the vulnerability, the larger the calculated safety cost factor value.

[0119] The cost factor of hoisted items is directly related to the value and fragility level of the hoisted items in the task attribute parameters. The higher the value and fragility of the items, the larger the calculated cost factor value of the hoisted items.

[0120] The equipment loss cost factor is calculated based on the cumulative working time, current load, and predicted remaining life of key components in the equipment status parameters. The longer the cumulative working time, the larger the current load, and the shorter the remaining life of the components, the larger the calculated equipment loss cost factor value.

[0121] The energy consumption cost factor is calculated based on the remaining battery power in the equipment status parameters, the current load, and the maximum allowable energy consumption in the task constraint parameters. The lower the remaining battery power, the higher the current load, and the smaller the allowable energy consumption margin, the larger the calculated energy consumption cost factor value.

[0122] The time cost factor is calculated based on the task deadline, current progress, and task criticality level in the task constraint parameters. The closer to the deadline, the more serious the progress lag, and the higher the task criticality level, the larger the calculated time cost factor value.

[0123] The predefined mapping rules are as follows:

[0124] The safety cost factor is mainly used to determine the initial value of the personnel risk weight coefficient, and the two are positively correlated. The hoisting item cost factor is mainly used to determine the initial value of the hoisting item risk weight coefficient, and the two are positively correlated. The equipment loss cost factor and the energy consumption cost factor are used together to determine the initial value of the equipment risk weight coefficient, and the initial value of the equipment risk weight coefficient is the result of the weighted sum of the two cost factors. The time cost factor is used to globally adjust the initial values ​​of the personnel risk weight coefficient, the equipment risk weight coefficient, and the hoisting item risk weight coefficient. When the time cost factor increases, the values ​​of all risk weight coefficients are reduced simultaneously while keeping the weight ratio of the three factors basically unchanged.

[0125] In this embodiment, the system collects four key parameters in real time: environmental parameters, task attributes, equipment status, and external constraints. Based on these parameters, it quantifies safety cost factors, hoisting item cost factors, equipment wear and tear cost factors, energy consumption cost factors, and time cost factors. Through predetermined mapping rules, safety costs are primarily mapped to personnel risk weights, hoisting item costs are mapped to hoisting item risk weights, and equipment wear and tear costs are jointly mapped to equipment risk weights. Time cost is used to globally adjust the weights of these three factors to reflect the impact of task urgency on overall risk tolerance. Finally, by normalizing the initial weights and using a first-order low-pass filter to achieve a smooth transition, the system outputs real-time dynamically adjusted risk weight coefficients for the three types of risks, summing to 1. This approach abandons the rigidity of traditional fixed-weight assessment methods and achieves intelligent and flexible shifts in the risk assessment focus. When the working environment is harsh and personal safety risks are prominent, the system automatically increases the personnel risk weight, making the warning more conservative; when the hoisting item is of high value, it focuses more on its integrity. This dynamic mapping based on real-time cost factors enables the early warning system to accurately respond to the priority changes of the core decision triangle of "safety, efficiency, and cost" under different operating conditions. At the same time, the weight smoothing mechanism effectively suppresses decision oscillations caused by instantaneous parameter fluctuations, ensuring the stability of control commands. Thus, in complex and ever-changing actual operations, it provides managers with dynamic risk assessment results that are more situationally aware and have greater decision support value.

[0126] Step 4: Combining the risk assessment results of the hoisted items, and comprehensively assessing personnel and equipment risks, a comprehensive risk index is generated by weighting and integrating the results based on determined dynamic weighting coefficients. This index specifically includes:

[0127] S41: Obtain the risk index of the hoisted items, and independently assess the personnel risk index and equipment risk index;

[0128] S42: Receive personnel risk weight coefficients, equipment risk weight coefficients, and hoisted item risk weight coefficients that are dynamically adjusted according to time and task status;

[0129] S43: The personnel risk index, equipment risk index, and hoisting item risk index are weighted and integrated with their corresponding dynamic weight coefficients to generate a comprehensive risk index;

[0130] S44: Compare the comprehensive risk index with the preset risk threshold to trigger different levels of risk warning signals. The higher the value of the comprehensive risk index, the higher the overall risk level of the system.

[0131] The personnel risk index is calculated based on the real-time location of personnel within the work area, the relative distance between the drone and personnel, and the current flight status parameters of the drone.

[0132] The equipment risk index is calculated based on the current power system status, structural stress status, and health status of the hoisting mechanism of the UAV.

[0133] In this embodiment, the present invention constructs an adaptive comprehensive risk assessment and early warning mechanism by integrating dynamically adjusted weighting coefficients with multi-dimensional risk indices. The technical solution is as follows: First, the system acquires independently assessed risk indices for hoisted items, personnel, and equipment indices in parallel; simultaneously, it receives a set of weighting coefficients that dynamically change with operational safety costs, economic costs, and time costs. Then, these three sub-risk indices are weighted and summed with their corresponding dynamic weights to calculate a single, quantitative comprehensive risk index. Finally, this index is compared with preset multi-level thresholds, thereby triggering early warning signals at different levels, from attention and warning to danger. This solution, by introducing dynamic weights, enables the assessment to automatically focus on the most significant risk issues, enhancing the pertinence and relevance of risk perception to decision-making; unifying multi-dimensional risks into a single index greatly simplifies the recognition and judgment of complex risks; and the tiered early warning mechanism transforms continuous risk values ​​into clear and actionable instructions, thus achieving a closed loop from quantitative risk assessment to tiered decision-making response, significantly improving the adaptability and overall safety of hoisting operation risk management.

[0134] Step 5: Based on historical monitoring data, control command sequences, and risk event records, a risk causal graph is constructed using a causal discovery algorithm to infer the root cause of the current risk state, specifically including:

[0135] S51: Collect multi-source time-series data within the historical time window, including sensor data sequences, control command sequences issued by the UAV flight control system, and recorded risk event label sequences. The length of the historical time window is dynamically adjusted according to the task complexity and data update frequency. The risk event label sequences include triggered risk warning levels, abnormal status indicators of hoisted items, and equipment fault codes.

[0136] S52: Based on multi-source time series data, a causal discovery algorithm is used to analyze and determine the causal relationship between various data variables. The causal discovery algorithm is one of the following: constraint-based causal discovery algorithm, score-based causal discovery algorithm, and time-delay-based causal discovery algorithm. Specifically, by testing the conditional independence between variables or optimizing the preset scoring function, the probability and causal direction of the existence of causal connection between variables are inferred, and connections that are only correlated but not causal are filtered out.

[0137] S53: Based on the established causal relationships, construct a directed graph model as a risk causal graph, where the nodes of the graph represent the variables of the multi-source time series data, and the directed edges between the nodes represent the direction of the causal relationship between the variables.

[0138] S54: Based on the constructed risk cause-effect graph, combined with current real-time monitoring data and status, infer the root cause nodes and key causal paths that led to the current risk status, specifically including:

[0139] S541: Evidence that maps current real-time monitoring data and system status to corresponding observed variable nodes and control variable nodes in a risk causal graph;

[0140] S542: In the risk causal graph, starting from the triggered risk variable node, trace back along the directed edge to identify all possible predecessor nodes as candidate cause nodes.

[0141] S543: Based on the causal strength and the degree to which real-time data deviates from the normal threshold, calculate the contribution of each candidate cause node to the current risk state;

[0142] S544: Based on the contribution ranking, the top-ranked candidate cause nodes and the paths connecting them to risk nodes are presumed to be the root cause and key causal path of the current risk.

[0143] In the risk-cause graph, the nodes are divided into three categories: observed variable nodes, control variable nodes, and risk variable nodes. The starting point of each directed edge is a dependent variable node, and the ending point is an effect variable node. Specifically:

[0144] The observed variable nodes correspond to sensor data sequences, including cable tension, hoisted object attitude, and environmental parameters;

[0145] The control variable node corresponds to the sequence of control commands, including commands for adjusting the speed, altitude, and attitude of the UAV.

[0146] The risk variable nodes correspond to a sequence of risk event labels, including the overall risk index and various sub-risk indices.

[0147] In this embodiment, the present invention constructs a dynamic knowledge graph capable of deeply analyzing the causes of risks by applying a causal discovery algorithm. Specifically, the system collects sensor data, control commands, and risk event records from historical operations, and automatically mines the causal relationships between observed variables, control variables, and risk variables using the causal discovery algorithm, constructing a structured risk causal graph. When a real-time operation triggers a risk warning, the system uses this graph as a basis to trace back along the causal edges from the risk node, and comprehensively calculates the contribution of each potential cause node, thereby accurately locating the most likely root cause and the complete risk-causing path. This method achieves a leap from risk phenomenon alarm to root cause and mechanism diagnosis, not only revealing the control logic or equipment correlation behind the risk, providing clear targets for precise intervention, but also continuously learning and optimizing the causal graph, enabling the system to possess deep analytical capabilities similar to expert experience, significantly improving the efficiency of operation and maintenance troubleshooting and the initiative of risk prevention.

[0148] Step Six: Using the obtained risk cause-and-effect diagram, extrapolate the preset control program and real-time input control commands to predict potential risks in the future time period, specifically including:

[0149] S61: Analyze the control programs currently to be executed and those to be executed in the future, capture real-time input control instructions, and convert them into the input sequence of the corresponding control variable nodes in the risk cause-effect graph, specifically:

[0150] Extract the sequence of control commands for the next N control cycles from the mission queue and command flow of the flight control system, including target spatial position, target attitude, desired velocity and desired acceleration; map the extracted discrete control commands to the trajectory of the value changes of each variable node representing the control input in the continuous prediction time domain in the risk cause-effect graph.

[0151] S62: Based on the risk cause-effect diagram and combined with the current system state, the impact of the control input sequence in the future time period is deduced along the direction of the causal relationship in the cause-effect diagram. This is achieved in the following way:

[0152] S621: Transform the constructed risk causal graph into a dynamic Bayesian network model, where the conditional probability distribution between nodes is learned from historical data;

[0153] S622: Input the real-time values ​​of the observed variables at the current moment and the predicted values ​​of the control variables at future moments as evidence into the dynamic Bayesian network model;

[0154] S623: Using a dynamic Bayesian network model for causal reasoning, calculate the posterior probability distribution of risk variable nodes at future time points given the current state and control sequence.

[0155] S63: Based on the simulation results, predict the probability of a specific risk event occurring in the future time period and output the risk prediction results.

[0156] The time step of the simulation is determined comprehensively based on the control frequency of the UAV, the dynamic response time of the controlled object, and the lead time required for early warning, and satisfies the following:

[0157] ;

[0158] in, For the time step of the simulation, To control the cycle, The average response time required from the issuance of a control command to a significant change in the system state. and An adjustment factor greater than 1 is used to provide a buffer margin for forecasting.

[0159] In this embodiment, the present invention achieves proactive prediction of operational risks by establishing a forward-looking risk simulation and deduction mechanism. Its core technology lies in transforming the control program to be executed (such as flight trajectory and speed commands) into an input sequence of control variables in a risk causal graph, and constructing a dynamic Bayesian network model based on this causal graph structure. By inputting the current real-time state and future control sequences into this model, the system can simulate the transmission effect of the "causal chain" after command execution, thereby calculating the probability distribution of various risk events occurring at different future times. Its beneficial effect is that this method advances risk warning from "post-event attribution" and "real-time alarm" to "pre-event prediction," enabling the system to identify high-risk control commands or flight phases before potential risks actually occur. This provides valuable decision-making buffer time for operators or autopilot systems to take adjustment, deceleration, or evasive measures, thereby greatly enhancing the proactive safety protection capability of hoisting operations.

[0160] Step Seven: Based on the comprehensive risk index and future risk forecast results, trigger the corresponding level of risk warning signal, and output control adjustment suggestions according to the warning level, including:

[0161] S71: Set multiple comprehensive risk index thresholds to classify risk warnings into multiple levels, including blue warning, yellow warning, orange warning and red warning;

[0162] S72: When the comprehensive risk index exceeds a certain level threshold, or the predicted probability of future risks exceeds a set threshold, the corresponding level of sound, light and electrical warning signals are triggered.

[0163] S73: When a yellow or higher warning is triggered, based on the root cause inferred from the risk cause-effect diagram and the predicted risk path, generate and output control adjustment suggestions, including adjusting flight speed, changing hoisting attitude, interrupting the current action, and executing emergency landing procedures.

[0164] In this embodiment, the invention first establishes four warning thresholds: blue, yellow, orange, and red. Based on the real-time comprehensive risk index and the predicted future risk probability, it dynamically triggers corresponding level sound, light, and electrical alarms. More importantly, when a yellow or higher warning is triggered, the system does not merely issue a warning; instead, it automatically generates and outputs specific control adjustment suggestions based on the root causes revealed by the risk cause-effect diagram and the predicted risk paths. This solution achieves seamless integration from risk quantification and prediction to action recommendations, directly transforming complex analysis results into actionable instructions. This significantly reduces the operator's cognitive load and decision-making delay, ensuring rapid and accurate execution from preventative adjustments for low-level warnings to decisive emergency responses for high-level warnings, thereby greatly improving the proactive safety and operational reliability of the entire hoisting operation system.

[0165] like Figure 2 As shown, this embodiment also provides a hoisting risk early warning system for hoisting drones, including:

[0166] The data acquisition module is used to monitor the stress data of multiple sets of hoisting cables and the attitude data of the hoisted items in the hoisting system in real time.

[0167] The risk analysis module for hoisted items is used to analyze the real-time stress state and stability of hoisted items based on the acquired stress and attitude data, and to assess the risk level of the hoisted items.

[0168] The dynamic weight calculation module is used to dynamically calculate the weight coefficients of personnel risk, equipment risk, and hoisted item risk based on the real-time operating environment and task parameters.

[0169] The comprehensive risk assessment module is used to integrate independently assessed personnel and equipment risks. It performs weighted fusion based on the dynamic weight coefficients determined by the dynamic weight calculation module to generate a comprehensive risk index.

[0170] The causal analysis and root cause inference module is used to construct a risk causal graph based on historical monitoring data, control command sequences and risk event records, and infer the root cause of the current risk status.

[0171] The risk prediction module is used to use a risk cause-effect diagram to extrapolate from preset control programs or real-time input control commands and predict potential risks in the future time period.

[0172] The early warning and response module is used to trigger risk warning signals of corresponding levels based on the comprehensive risk index and future risk prediction results, and output control adjustment suggestions according to the warning level.

[0173] Example:

[0174] In a high-rise building photovoltaic system installation project, a drone was required to lift a valuable and fragile photovoltaic glass panel, 2.5 meters long and 1.2 meters wide, to the rooftop. Force sensors on four lifting cables synchronously collected data at a 100Hz frequency, displaying real-time tension fluctuations between 450N and 520N. An IMU installed at the center of the glass panel monitored its roll angle, which periodically oscillated between -3 degrees and +5 degrees, with an angular velocity of approximately 0.8 degrees per second. Continuous measurements by a lidar sensor on the drone's underside showed that the relative distance between the glass panel and the drone changed by a minute ±0.15 meters under slight wind disturbance.

[0175] Based on the above data, the system calculated the force imbalance parameter to be 0.15. Short-time Fourier transform analysis of the IMU's acceleration data revealed a weak resonant energy peak close to the cable's oscillation frequency. The attitude stability parameter was calculated to be 2.4. After matching the feature vector with a preset pattern library, the system identified the current dominant instability mode as "torque imbalance-dominated rotational instability" and calculated the risk factor for the hoisted item to be 0.28. Because photovoltaic glass is a brittle material and the instantaneous wind speed reached 8 m / s, the system dynamically lowered the risk level threshold based on material properties and environmental parameters.

[0176] The dynamic weight calculation module considers the current operating conditions: wind speed 8 m / s, operating height 80 m, high value and fragility of photovoltaic glass, cumulative drone motor operating time 120 hours, remaining battery power 65%, and task completion within 30 minutes. Calculations show a significant increase in safety cost factors and hoisting item cost factors, while the time cost factor is moderate. Based on this, the system dynamically allocates weights: personnel risk weight 0.3, equipment risk weight 0.2, and hoisting item risk weight 0.5. After a first-order low-pass filter, the weights smoothly transition to the new values.

[0177] The comprehensive risk assessment module received the following data: risk index of the hoisted item (0.28), risk index of surrounding personnel calculated via UWB positioning (0.1), and equipment risk index based on motor temperature and vibration analysis (0.15). After dynamic weighted fusion, the current comprehensive risk index is 0.216, and the system classifies it as a "yellow alert" level.

[0178] Meanwhile, the causal analysis and root cause inference module is running continuously. Based on historical data from the past week, the system has constructed a risk causal graph using a causal discovery algorithm. Nodes in the graph include cable tension, attitude angle, wind speed, and control commands. Once the current risk of rotational instability is identified, the system traces back along the causal graph and, combined with real-time data, infers the root cause nodes as "a sudden drop in tension on cable #3" and "a minor yaw acceleration command issued by the UAV 10 seconds ago." The key causal path clearly shows how the control command affects the UAV's attitude, ultimately causing the load to oscillate.

[0179] The risk prediction module immediately intervened. It extracted the planned instruction sequence for the next 15 seconds from the flight control system queue and input it into a dynamic Bayesian network built on a risk causal graph for simulation. The simulation showed that if the preset acceleration instructions planned to counteract wind disturbances continued to be executed, the probability of increased load sway would exceed 70%, and the comprehensive risk index might rise above 0.5 after 8 seconds, triggering an orange alert.

[0180] The warning and response module was immediately triggered. The system activated a yellow audible and visual alarm signal on the control terminal. Simultaneously, specific control adjustment suggestions popped up on the screen: "A tendency towards rotational instability has been detected, primarily related to recent yaw commands. Continuing the current trajectory is predicted to increase the risk. Recommendations: 1. Immediately reduce flight speed by 30%; 2. Suspend the planned acceleration command and activate anti-sway control mode; 3. Adjust the sling attitude, making a slight adjustment of 0.5 degrees towards cable #3." The operator adopted the suggestions, and after the system executed them, real-time monitoring data showed that the sway amplitude gradually converged, the overall risk index steadily decreased to a safe range, and the warning was lifted.

[0181] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for early warning of lifting risks for hoisting drones, characterized in that: Includes the following steps: S1: Real-time monitoring of the force data of multiple sets of hoisting cables and the attitude data of the hoisted items in the hoisting system; S2: Based on the acquired force and attitude data, analyze the real-time force state and stability of the hoisted item, and assess the risk level of the hoisted item; S3: Dynamically calculate the weighting coefficients of personnel risk, equipment risk, and hoisted item risk based on the real-time operating environment and task parameters; S4: Combining the risk assessment results of the hoisted items, and comprehensively assessing personnel and equipment risks, a comprehensive risk index is generated by weighting and integrating the results based on the determined dynamic weight coefficients. S5: Based on historical monitoring data, control command sequences, and risk event records, a risk causal graph is constructed using a causal discovery algorithm to infer the root cause of the current risk status; S6: Using the obtained risk cause-effect diagram, extrapolate the preset control program and real-time input control commands to predict potential risks in the future time period; S7: Based on the comprehensive risk index and future risk prediction results, trigger the corresponding level of risk warning signal, and output control adjustment suggestions according to the warning level; Specifically, assessing the risk level of hoisted items includes: S21: Based on the real-time force data of each hoisting cable, calculate the resultant force, resultant moment and force imbalance parameters of the hoisted item; S22: Based on the acquired attitude data of the hoisted item, analyze the vibration characteristics, attitude stability and torsional angle parameters of the hoisted item; S23: Couple the stress imbalance parameter with the attitude stability parameter to identify the instability mode of the hoisted item; S24: Based on the instability mode, a risk quantification model is used to calculate the risk factors of the hoisted items; S25: Based on the material properties, structural characteristics, and current environmental conditions of the hoisted items, dynamically adjust the risk level threshold range, and judge the risk factors of the hoisted items according to the risk level threshold range to determine the specific risk level; Specifically, when coupling the stress imbalance parameter with the attitude stability parameter to identify the instability mode of the hoisted object, the analysis includes: S231: Combine the force imbalance parameter, the reciprocal of the attitude stability parameter, the proportion of dangerous vibration energy, and the torsion angle parameter to construct a coupled analysis feature vector; S232: Calculate the matching degree between the coupling analysis feature vector and a set of preset instability mode feature vectors respectively; S233: Based on the calculated matching degree, determine the instability mode with the highest matching degree to the current state and identify it as the current main instability mode; When dynamically calculating the weighting coefficients for personnel risk, equipment risk, and hoisted item risk based on real-time operating environment and task parameters, the specific calculations include: S31: Real-time acquisition of environmental parameters for the current operation, task attribute parameters for the hoisting task, status parameters of the UAV equipment, and task constraint parameters from external systems; S32: Calculate the safety cost factor, hoisting cost factor, equipment loss cost factor, energy consumption cost factor, and time cost factor based on various parameters. S33: Convert the safety cost factor, hoisting item cost factor, equipment wear and tear cost factor, energy consumption cost factor, and time cost factor into the initial values ​​of personnel risk weight coefficient, equipment risk weight coefficient, and hoisting item risk weight coefficient through a predetermined mapping rule; S34: Normalize the initial values ​​of personnel risk weight coefficient, equipment risk weight coefficient, and hoisted item risk weight coefficient, and output them as the final dynamic weight coefficients.

2. The lifting risk early warning method for lifting drones according to claim 1, characterized in that: Real-time monitoring of the stress data of multiple sets of lifting cables and the attitude data of the lifted items in the lifting system specifically includes: S11: Force sensors are installed on each hoisting cable to synchronously collect tension data of each cable at a sampling frequency of 100Hz. S12: Install an inertial measurement unit on the hoisted item to collect the acceleration, angular velocity and attitude angle data of the hoisted item in real time; S13: By installing a ranging module on the drone, the relative position and distance between the hoisted object and the drone body are monitored.

3. The lifting risk early warning method for lifting drones according to claim 2, characterized in that: The pre-defined mapping rules are as follows: The safety cost factor is mainly used to determine the initial value of the personnel risk weight coefficient, and the two are positively correlated. The hoisting item cost factor is mainly used to determine the initial value of the hoisting item risk weight coefficient, and the two are positively correlated. The equipment loss cost factor and the energy consumption cost factor are used together to determine the initial value of the equipment risk weight coefficient, and the initial value of the equipment risk weight coefficient is the result of the weighted sum of the two cost factors. The time cost factor is used to globally adjust the initial values ​​of the personnel risk weight coefficient, the equipment risk weight coefficient, and the hoisting item risk weight coefficient. When the time cost factor increases, the values ​​of all risk weight coefficients are reduced simultaneously while keeping the weight ratio of the three factors basically unchanged.

4. The lifting risk early warning method for lifting drones according to claim 3, characterized in that: When constructing a risk causal graph using a causal discovery algorithm based on historical monitoring data, control command sequences, and risk event records to infer the root cause of the current risk state, the specific steps include: S51: Collect multi-source time-series data within the historical time window; S52: Based on multi-source time series data, a causal discovery algorithm is used to analyze and determine the causal relationships between various data variables; S53: Based on the established causal relationships, construct a directed graph model as a risk causal graph; S54: Based on the constructed risk causal graph, combined with the current real-time monitoring data and status, infer the root cause node and key causal path that caused the current risk status.

5. A method for early warning of lifting risks for hoisting drones according to claim 4, characterized in that: When inferring the root cause nodes and key causal paths leading to the current risk state based on the constructed risk cause-effect graph and combined with current real-time monitoring data and status, the specific steps include: S541: Evidence that maps current real-time monitoring data and system status to corresponding observed variable nodes and control variable nodes in a risk causal graph; S542: In the risk causal graph, starting from the triggered risk variable node, trace back along the directed edge to identify all possible predecessor nodes as candidate cause nodes. S543: Based on the causal strength and the degree to which real-time data deviates from the normal threshold, calculate the contribution of each candidate cause node to the current risk state; S544: Based on the contribution ranking, the top-ranked candidate cause nodes and the paths connecting them to risk nodes are presumed to be the root cause and key causal path of the current risk.

6. A method for early warning of lifting risks for hoisting drones according to claim 5, characterized in that: When using the obtained risk cause-effect diagram to extrapolate the preset control program and real-time input control commands to predict potential risks in the future time period, the specific steps include: S61: Analyze the control programs that are currently to be executed and those that will be executed in the future, and capture the control instructions that are input in real time, and convert them into the input sequence of the corresponding control variable node in the risk cause-effect graph; S62: Based on the risk cause-effect diagram and combined with the current system state, extrapolate the impact of the control input sequence in the future time period along the direction of the causal relationship in the cause-effect diagram; S63: Based on the simulation results, predict the probability of a specific risk event occurring in the future time period and output the risk prediction results.

7. A hoisting risk early warning system for hoisting drones, used to implement the hoisting risk early warning method for hoisting drones as described in any one of claims 1-6, characterized in that: include: The data acquisition module is used to monitor the stress data of multiple sets of hoisting cables and the attitude data of the hoisted items in the hoisting system in real time. The risk analysis module for hoisted items is used to analyze the real-time stress state and stability of hoisted items based on the acquired stress and attitude data, and to assess the risk level of the hoisted items. The dynamic weight calculation module is used to dynamically calculate the weight coefficients of personnel risk, equipment risk, and hoisted item risk based on the real-time operating environment and task parameters. The comprehensive risk assessment module is used to integrate independently assessed personnel and equipment risks. It performs weighted fusion based on the dynamic weight coefficients determined by the dynamic weight calculation module to generate a comprehensive risk index. The causal analysis and root cause inference module is used to construct a risk causal graph based on historical monitoring data, control command sequences and risk event records, and infer the root cause of the current risk status. The risk prediction module is used to use a risk cause-effect diagram to extrapolate from preset control programs or real-time input control commands and predict potential risks in the future time period. The early warning and response module is used to trigger risk warning signals of corresponding levels based on the comprehensive risk index and future risk prediction results, and output control adjustment suggestions according to the warning level.

Images